Detection and characterization of critical under-immunized hotspots Emergence of undervaccinated geographical clusters for diseases like measles has become a national concern. A number of measles outbreaks have occurred in recent months, despite high MMR coverage in the United States ( 95%). Such undervaccinated clusters can act as reservoirs of infection that can transmit the disease to a wider population, magnifying their importance far beyond what their absolute numbers might indicate. The existence and growth of such undervaccinated clusters is often known to public health agencies and health provider networks, but they typically do not have enough resources to target people in each such cluster, to attempt to improve the vaccination rate. Preliminary results show that not all undervaccinated clusters are ?equal? in terms of their potential for causing a big outbreak (referred to as its ?criticality?), and the rate of undervaccination in a cluster does not necessarily correlate with its criticality. However, there are no existing methods to estimate the potential risk of such clusters, and to identify the most ?critical? ones. Some of the key reasons are: (i) purely data-driven spatial statistics methods rely only on immunization coverage, which does not give any indication of the risk of an outbreak; and (ii) current causal epidemic models need to be combined with detailed incidence data, which has not been easily available. This proposal brings together a systems science approach, combining agent-based stochastic epidemic models, and techniques from machine learning, high performance computing, data mining, and spatial statistics, along with novel public and private datasets on immunization and incidence, to develop a novel methodology for identifying critical clusters, through the following tasks: (i) Identify spatial clusters with signi?cantly low immunization rates, or strong anti-vaccine sentiment; (ii) Develop an agent based model for the spread of measles that incorporates detailed immunization data, and is calibrated using a novel source of incidence data; (iii) Develop methods to ?nd and characterize critical spatial clusters, with respect to different metrics, which capture both epidemic and economic burden, and order underimmunized clusters based on their criticality; and (iv) Use the methodology to evaluate interventions in terms of their effect on criticality. A highly interdisciplinary team involving two universities, a health care delivery organization and a state department of Health, will work together to develop this methodology. Characterization of such clusters will enable public health departments and policy makers in targeted surveillance of their regions and a more ef?cient allocation of resources.